Logging parameters and metrics in MLOps - Practice Problems & Coding Challenges

Challenge - 5 Problems

💻 Command Output

intermediate

What is the output of this MLflow logging command?

Consider the following Python code snippet using MLflow to log parameters and metrics. What will be printed after running this code?

MLOps

import mlflow
with mlflow.start_run():
    mlflow.log_param("learning_rate", 0.01)
    mlflow.log_metric("accuracy", 0.95)
    print(mlflow.active_run().info.run_id)

A A valid UUID string representing the run ID

B None

C SyntaxError

D An empty string

💡 Hint

mlflow.start_run() creates a new run and mlflow.active_run() returns the current run info.

The mlflow.start_run() context manager starts a new run. mlflow.active_run() returns the current run object, and its info.run_id is a unique identifier string (UUID) for the run.

🧠 Conceptual

intermediate

Which MLflow function logs a hyperparameter for a model training run?

You want to record the value of a hyperparameter named 'batch_size' with value 32 during your ML experiment. Which MLflow function should you use?

A mlflow.log_param('batch_size', 32)

B mlflow.log_metric('batch_size', 32)

C mlflow.log_artifact('batch_size', 32)

D mlflow.set_tag('batch_size', 32)

💡 Hint

Parameters are fixed values describing the run, metrics are numeric results that can change over time.

mlflow.log_param() is used to log parameters like hyperparameters. mlflow.log_metric() is for numeric results like accuracy. mlflow.log_artifact() is for files. mlflow.set_tag() adds metadata tags.

❓ Troubleshoot

advanced

Why does this MLflow metric logging code fail to record the metric?

You run this code but the metric 'loss' does not appear in the MLflow UI after the run completes. What is the most likely reason? import mlflow mlflow.log_metric('loss', 0.25)

A The metric name 'loss' is invalid and causes an error

B The metric value 0.25 is not a valid float

C MLflow requires metrics to be logged only after the run ends

D No active MLflow run was started before logging the metric

💡 Hint

MLflow needs a run context to associate logged data with.

MLflow requires an active run context (started with mlflow.start_run()) to log parameters or metrics. Without it, the log_metric call does nothing.

🔀 Workflow

advanced

What is the correct sequence to log parameters and metrics in MLflow during a training run?

Arrange the steps in the correct order to log parameters and metrics properly in MLflow.

A 1,3,2,4

B 2,1,3,4

C 1,2,3,4

D 3,2,1,4

💡 Hint

You must start the run before logging anything.

You first start the run, then log parameters before training, then train and compute metrics, then log metrics.

✅ Best Practice

expert

Which practice ensures reliable metric logging in distributed training with MLflow?

In a distributed training setup with multiple workers, which approach best ensures metrics are logged correctly without duplication or loss?

A All workers log metrics independently during training

B Only the main worker logs metrics after aggregating results from all workers

C Metrics are logged only before training starts

D Each worker logs metrics with the same run ID simultaneously

💡 Hint

Avoid multiple workers writing to the same run at the same time.

In distributed training, only the main worker should log aggregated metrics to avoid duplication and conflicts. Other workers should not log metrics independently.